Based On Multi-images To Reconstruct Eave Tiles’s 3D Model Algorithm Research And Implement

Eave tiles from Qin and Han dynasties is important carriers for researchers to study cultural, political, economic and other information of ancient times. It has very important historical values. The rapid rise of the digitization of cultural heritage protection and promotion of 3D printing provide new techniques for the restoration, protection and copy of eave tiles, which are based on fast 3D reconstruction of eave tiles. Addressing the problem of the lack of 3d models of eave tiles, we use multiple images to reconstruct the 3d model of eave tiles. This paper does those works as following:1. According to the special structure of eave tiles, SIFT feature extraction algorithm with the local entropy constraint is better than other feature extraction algorithm.2. Because the original PMVS algorithm’s initial patch is too sparse to expand, we propose an improved PMVS algorithm with quasi dense matching to increase the number of initial patches, which could reduce the work of expanding process. The result of experiment proves that the improved PMVS algorithm has higher quality and more details than the original algorithm.3. Fundamental matrix is used to take wrong points out. And Poisson process is applied to reconstruct the surface and to repair the holes.4. The management system of eave tiles from Qin and Han is designed and established based on the experiment, which includes four functions:data administration, eave tile demonstration,3D reconstruction and character recognition.